DEV Community

Cover image for I open-sourced the cockpit we used to understand huge engineering departments in days
yaniv zalevas
yaniv zalevas

Posted on

I open-sourced the cockpit we used to understand huge engineering departments in days

#productivity #devops #opensource #showdev

My name is Yaniv, and I run Codpal, a CTO-as-a-service company. Over the years we've walked into a lot of new companies, and we always needed the same thing: a fast understanding of how engineering actually runs — what's in the roadmap, what the developers are focused on, who the main developers carrying the work really are.

Every engagement started with weeks of meetings to map what the data could have told us in hours.

So we built the tool

Deckgauge pulls Jira, GitHub, GitLab and Azure DevOps into a single data warehouse of development information, and gives you the cockpit picture:

  • DORA metrics — deployment frequency, lead time, change failure rate, time to restore, computed from your actual delivery data
  • Team Focus — what every team is actually working on vs. what the roadmap says
  • Per-engineer workload — who's carrying the delivery, who's drowning
  • CapEx vs OpEx — how much of engineering is building the future vs. keeping the lights on
  • One board over all four trackers, with org-chart views

It let us sit with division leads on day one and speak their language — with real data instead of guesses.

The part you might find interesting

Deckgauge was largely built by AI agents (Claude Code with an agent workflow) — human-reviewed on every merge. The tool that measures engineering is itself an experiment in AI-built engineering. Scroll the commit history; it's the artifact.

As featured on YouTube

The Next New Thing (56K subscribers), a weekly show covering the best GitHub repos, featured Deckgauge — demoing the board live and taking on the "why not just Jira?" question:

🎬 https://youtu.be/1fHsIveXRa8?t=1445

We open-sourced it

It worked so well we decided it's a gift for everyone to use:

Fully self-hosted — single Docker deploy, nothing leaves your infrastructure. In a world where every engineering-metrics SaaS wants your ticket data in their cloud, we think that matters. License is FSL-1.1 (free for any use except offering it as a competing product; converts to Apache-2.0 in September 2028).

If you're an engineering leader who wants visibility without the vendor cloud, or you just want to see what an AI-agent-built codebase looks like — take it for a spin. Feedback welcome, especially the critical kind.

Cross-post note: this continues the AI-agents journey from my earlier posts here — the tool that measures engineering is itself an experiment in AI-built engineering.

Top comments (0)